An evolutionary approach for worst-case tolerance design

摘要

jttsai@mail.npue.edu.tw, jinntsong.tsai@gmail.com; ;A sliding-Ievel orthogonal differential evolution algorithm with a two-level orthogonal array (SLO-DEA20A) is proposed for solving worst-case tolerance design problems. Tolerance affects system performance and leads to violate design constraints. By including a two-level orthogonal array, the proposed SL0DEA20A obtains robust optimal solutions that minimize the impact of parameter variations and that maintain compliance with a comprehensive constraint set. Two design examples are used for performance evaluation of the SL0DEA20A. The first is a 10-variable function, which includes linear, non-linear, quadratic, and polynomial forms to illustrate its general robustness and computational efficiency. The second example is a speed reducer design that involves seven variables and multiple non-linear engineering constraints. The SL0DEA20A is also compared with sliding-Ievel orthogonal differential evolution algorithms with either three-level orthogonal array or two-level full-factorial design. Additionally, performance comparisons confirm that the proposed SL0DEA20A outperforms nature-inspired methods presented in the literature.